set	n	unit	defining_filter	used_for
Raw phylostratigraphic candidates	7252	genes (9 assemblies)	restricted at or below the base of Anophelinae; HDF-suspect loci excluded where the Weisman model was computable (see Methods 4.2)	starting pool
Cross-species protein families	5359	families	DIAMOND all-vs-all, single-linkage clustering of the raw candidates	clustering unit
Multi-member families	877	families	more than one member gene (includes within-assembly paralogs)	Table 1 stage 3
Multi-assembly families	800	families	present in more than one assembly (the artifact-relevant subset of the above)	artifact reasoning
Complex-core families	213	families	present in >=3 species of the gambiae complex; all nine focal assemblies are complex members, so this equals presence in >=3 species	all downstream conclusions
Core-family member genes	1311	genes	all member genes of the 213 complex-core families	TE screen denominator
Uncharacterized core representatives	118	families	complex-core families with no characterized product	domainless screen (Table 1 stage 5)
Domainless core families	117	families	uncharacterized minus TRGfam00107 (PF10642:Tom5), the only one with a Pfam domain; = 99.2% of 118	domainlessness statistic
Domainless de novo novelty set	116	families	domainless minus TRGfam00131 (diverged Rho-GEF, phylostratigraphy false positive)	PRIMARY de novo set (Table 1 stage 6)
"""Dark"" families (no SwissProt homolog)"	116	families	uncharacterized minus the two SwissProt hits (TRGfam00131, TRGfam00189). PARALLEL property of the 118, not a subset of the novelty set: the two sets intersect at 115	database-coverage control
De novo-containing core families	47	families	complex-core families with >=1 member gene classified de novo at the GENE level (distinct from the family-level novelty set above)	TE screen (Section 2.2)
AlphaFold3-modelled core reps	209	families	complex-core families <1,000 aa (of 213); + 20 housekeeping positive controls	structural analysis
Modelled de novo families	115	families	novelty set minus 2 families >=1,000 aa (TRGfam00131 and TRGfam00184, not modelled) and minus TRGfam00107	structural analysis
Confident de novo models	27	models	modelled de novo with mean pLDDT >=70; FoldSeek vs PDB100 (E<1e-3) recovered 0 known folds	PRIMARY structural result (0/27)
NTic (synteny de novo)	24	families	synteny classifier majority = NTic (non-coding-derived)	origin evidence
AAic (synteny diverged duplicate)	53	families	synteny classifier majority = AAic (aligns to annotated coding)	counted separately, NOT de novo
Unknown (synteny unresolved)	136	families	no classifiable syntenic block could be aligned	resolved partly by DENSE
DENSE genome-wide de novo (PEST)	152	of 400 candidates	DENSE non-coding homology search, An. gambiae PEST	orthogonal confirmation
High-confidence de novo	37	An. gambiae genes	NTic AND DENSE-confirmed	structural analysis anchors
Complex-wide de novo TRG genes	1335	genes	trg_class = de_novo_TRG across all nine assemblies (gene-level, all strata)	catalog scope